【摘要】 考虑联轴器不对中、转定子碰摩、非线性油膜力及非线性弹簧力等因素，建立了以滑动轴承为支撑的多故障耦合转子动力学模型。针对润滑油粘度对轴承支撑力的影响，引入粘度系数进行系统数值求解，获得转盘径向位移作为输出响应值。采用河马（HO）优化算法对反向传播（BP）神经网络进行优化，将响应值数据分别输入两种网络进行训练，并对比其预测误差。利用训练好的优化模型对10 000组响应值进行预测，结合蒙特卡洛法分别计算数值求解与网络预测所得响应值的可靠度。结果表明：经HO算法优化的BP模型具有更强的收敛性能；两种方法所得可靠度误差为0.18%，满足可靠度精度要求，有效扩大了蒙特卡洛法求解可靠度的样本量。更多还原

【Abstract】 Considering factors such as misalignment of couplings,friction between the rotor and stator,nonlinear oil film force and nonlinear spring force,a multi-fault coupled rotor dynamics model supported by sliding bearings is established. Aiming at the influence of lubricating oil viscosity on the bearing support force,the viscosity coefficient is introduced for systematic numerical solution,and the radial displacement of the turntable is obtained as the output response value. The hippopotamus optimization(HO) algorithm is adopted to optimize the back propagation(BP)neural network. The response value data are input into the two networks respectively for training,and their prediction errors are compared. The trained optimization model is used to predict 10 000 sets of response values,and the reliability of the response values obtained by numerical solution and network prediction is calculated respectively in combination with the Monte Carlo method. The results show that the BP model optimized by the HO algorithm has stronger convergence performance. The reliability errors obtained by the two methods are 0. 18%,which meets the reliability accuracy requirements and effectively expands the sample size for solving reliability by the Monte Carlo method.更多还原